The paper proposes Augmented Concept Activation Vector (ACAV), which injects a visual concept into input images and measures the resulting activation shift to quantify that concept's influence on a classifier's decision.
Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Developing Explainable Machine Learning Model using Augmented Concept Activation Vector
The paper proposes Augmented Concept Activation Vector (ACAV), which injects a visual concept into input images and measures the resulting activation shift to quantify that concept's influence on a classifier's decision.